Abstract:

To address the issue of low approval accuracy of current recommendation systems, a proposed algorithm incorporating time features and multisimilarity is made to enhance the effect of longstanding data, handler awareness, and mission prominence on the recommendation algorithm. Additionally, the likeness of different types of users is familiarized to enhance the issue of unfriendly starts to a convincing extent. Time is added to the algorithm as a scale factor because the longer it takes, the lower the likelihood of selection. To avoid misjudging high-scoring and popular items, we normalize the popularity when the behavior takes place, i.e., attention in the mission, to give justice to similar users rather than the score value. New users lack historical score records.

Description:

.